An Updated Hindcast of Arctic September Sea Ice Extent Using Teleconnection Indices

The decline of Arctic sea ice in recent decades has prompted efforts to hindcast sea ice extent using large scale teleconnection patterns. In a recent study, regression models were used to predict September sea ice extent using teleconnection indices between 1948 and 2000. In the current work, we extend the study to 2020, with hindcast regression models focused on the influence of several monthly teleconnection indices, including the Arctic Oscillation (AO), the Central Arctic Index (CAI; i.e., a practical version of Arctic Dipole Anomaly, DA), the North Atlantic Oscillation (NAO), the El Niño-Southern Oscillation (ENSO), the Atlantic Multidecadal Oscillation (AMO), and the Pacific Decadal Oscillation (PDO). Hindcast regression models were developed such that only the most statistically significant teleconnection patterns were considered (p less than or equal to 0.10), improving the computational efficiency compared to previous model iterations. The extended time frame of the current study (1950 – 2020) improved model performance as compared to the original study (1948 – 2000). Teleconnection influence was examined through pre-training and post-training exclusions of teleconnection indices, with both methods demonstrating similar results. AMO and AO were consistently the biggest influence on Arctic sea ice extent, while ENSO and PDO were the smallest. CAI and NAO were often varied in their influence, ranging from negligible to moderate with respect to the other teleconnections.